Skin Strain Measurement Technique
Step 2: Feature Extraction
Online fingerprint recognition generally relies on two types of salient features of the fingerprint called minutiae: ridge endings and ridge bifurcations. Pores, which appear as small openings on the surface of the fingerprint ridges, are also used occasionally. Here we are interested in extracting three types of features: valley endings, valley bifurcations, and pores.
The feature extraction process uses a number of image processing operations to simplify, enhance and analyze the fingerprint images. The full process is illustrated in Figure 1.
Figure 1. Feature extraction block diagram.
Figure 2 illustrates some of the operations performed on a small segment of a fingerprint. The original fingerprint consists of a grayscale image with white pores and valleys. This image is binarized to indicate which pixels belong to valleys and pores. A connected-component analysis then extracts many pores in the image. The binary fingerprint, minus pores, is then thinned to simplify processing. The resulting skeleton is analyzed to locate valley features. Syntactic rules are finally applied to correct some common artifacts. Notice that valley features have an associated direction which can be used for matching.
|
|
|
|
|
(a) Grayscale fingerprint. |
(b) Binary fingerprint. |
(c) Grayscale pores. |
|
|
|
|
|
(d) Thinned fingerprint with pores. |
(e) Extracted features with minutiae orientation (before corrections). |
(f) Extracted features with minutiae orientation (after corrections). |
Figure 2. Feature extraction in fingerprint segment.